Supervised Learning-based Receptor Abundance Estimation using STREAK: An Application to the 10X Genomics human extranodal marginal zone B-cell tumor/mucosa-associated lymphoid tissue (MALT) dataset

Load the STREAK package

STREAK is a supervised receptor abundance estimation method that depends on functionalities from the Seurat (Hao et al. 2021; Stuart et al. 2019; Butler et al. 2018; Satija et al. 2015), SPECK (Frost and Javaid 2022), VAM (Frost 2021) and Ckmeans.1d.dp (Wang and Song 2011; Song and Zhong 2020) packages.

library(STREAK)

Receptor gene set construction using a subset of joint scRNA-seq/CITE-seq training data

STREAK performs receptor abundance estimation by leveraging expression associations learned from joint scRNA-seq/CITE-seq training data. These associations can either be manually specified using pre-existing ground truth or can be built using a subset of joint transcriptomics and proteomics data. Below, we use a subset of 1000 cells from the 10X Genomics human extranodal marginal zone B-cell tumor/mucosa-associated lymphoid tissue (MALT) scRNA-seq/CITE-seq joint dataset to build a gene set weights membership matrix for the CD3, CD4, CD8a, CD14 and CD15 receptors. Given a \(m \times n\) training scRNA-seq counts matrix and a \(m \times h\) CITE-seq matrix, the receptorGeneSetConstruction() function is utilized to learn associations between each CITE-seq ADT transcript and all scRNA-seq transcripts. The resulting gene weights membership matrix is \(n \times h\).

data("train.malt.rna.mat")
data("train.malt.adt.mat")
receptor.geneset.matrix.out <- receptorGeneSetConstruction(train.rnaseq = 
                                                  train.malt.rna.mat, 
                                                train.citeseq = 
                                                  train.malt.adt.mat[,1:5], 
                                                rank.range.end = 100, 
                                                min.consec.diff = 0.01, 
                                                rep.consec.diff = 2,
                                                manual.rank = NULL, 
                                                seed.rsvd = 1)
dim(receptor.geneset.matrix.out)
#> [1] 33538     5
head(receptor.geneset.matrix.out)
#>                      CD3         CD4        CD8a        CD14        CD15
#> MIR1302-2HG -0.604110407 -0.26998235  0.27026601  0.57776984  0.57784085
#> FAM138A     -0.598787274 -0.25872687  0.09013295  0.57990947  0.58231208
#> OR4F5       -0.095207766 -0.20187439  0.15526743  0.05700447  0.15424909
#> AL627309.1   0.067616000  0.08191478 -0.09400833 -0.09638293 -0.06789212
#> AL627309.3  -0.008013746  0.13915141 -0.14105265  0.01494577 -0.02837786
#> AL627309.2  -0.098468902 -0.06671758  0.01694013  0.08150649  0.08697406

Receptor abundance estimation for target scRNA-seq data

Following the development of weighted gene sets, the receptorAbundanceEstimation() function is used to perform receptor abundance estimation. A subset of 1100 cells from the 10X Genomics MALT scRNA-seq data is used for estimation. Given a \(m \times n\) target scRNA-seq counts matrix and a \(n \times h\) gene set weights membership matrix, target scRNA-seq expression from top most weighted genes with each ADT transcript is used for gene set scoring and subsequent thresholding. The resulting estimated receptor abundance matrix is \(m \times h\).

data("target.malt.rna.mat")
receptor.abundance.estimates.out <- 
  receptorAbundanceEstimation(target.rnaseq = target.malt.rna.mat,
                              receptor.geneset.matrix = 
                                receptor.geneset.matrix.out,
                              num.genes = 10, rank.range.end = 100, 
                              min.consec.diff = 0.01, rep.consec.diff = 2,
                              manual.rank = NULL, seed.rsvd = 1, 
                              max.num.clusters = 4, seed.ckmeans = 2)
dim(receptor.abundance.estimates.out)
#> [1] 1100    5
head(receptor.abundance.estimates.out)
#>                          CD3 CD4      CD8a      CD14      CD15
#> CTACCTGAGAGCGACT-1 0.0000000   0 0.9987944 0.6740526 0.7753415
#> TGGCGTGCACAGCATT-1 0.9464793   0 0.0000000 0.0000000 0.0000000
#> TAGGAGGAGCTGGCCT-1 0.0000000   0 0.0000000 0.9992784 0.9988085
#> ACTATCTCACCCTATC-1 0.0000000   0 0.9982689 0.1559718 0.2513592
#> ACGGAAGTCAATCCGA-1 0.0000000   0 0.9957439 0.5229880 0.6813975
#> AAGTACCCACAGAGCA-1 0.0000000   0 0.0000000 0.9990658 0.9985386

References

Butler, Andrew, Paul Hoffman, Peter Smibert, Efthymia Papalexi, and Rahul Satija. 2018. “Integrating Single-Cell Transcriptomic Data Across Different Conditions, Technologies, and Species.” Nature Biotechnology 36: 411–20. https://doi.org/10.1038/nbt.4096.

Frost, H. Robert. 2021. VAM: Variance-Adjusted Mahalanobis. https://CRAN.R-project.org/package=VAM.

Frost, H. Robert, and Azka Javaid. 2022. SPECK: Receptor Abundance Estimation Using Reduced Rank Reconstruction and Clustered Thresholding. https://CRAN.R-project.org/package=SPECK.

Hao, Yuhan, Stephanie Hao, Erica Andersen-Nissen, et al. 2021. “Integrated Analysis of Multimodal Single-Cell Data.” Cell, ahead of print. https://doi.org/10.1016/j.cell.2021.04.048.

Satija, Rahul, Jeffrey A Farrell, David Gennert, Alexander F Schier, and Aviv Regev. 2015. “Spatial Reconstruction of Single-Cell Gene Expression Data.” Nature Biotechnology 33: 495–502. https://doi.org/10.1038/nbt.3192.

Song, Mingzhou, and Hua Zhong. 2020. “Efficient Weighted Univariate Clustering Maps Outstanding Dysregulated Genomic Zones in Human Cancers.” Bioinformatics 36 (20): 5027–36. https://doi.org/10.1093/bioinformatics/btaa613.

Stuart, Tim, Andrew Butler, Paul Hoffman, et al. 2019. “Comprehensive Integration of Single-Cell Data.” Cell 177: 1888–902. https://doi.org/10.1016/j.cell.2019.05.031.

Wang, Haizhou, and Mingzhou Song. 2011. “Ckmeans.1d.dp: Optimal \(k\)-Means Clustering in One Dimension by Dynamic Programming.” The R Journal 3 (2): 29–33. https://doi.org/10.32614/RJ-2011-015.